Research On Radar Target Data Association Method Based On Classification Of Measurement Data

The maneuverability of the target and the complexity of the environment greatly increase the difficulty of the target data association and corresponding trajectory formation in radar data processing system.The conventional target association methods rely too much on prior information and has inferior performance on data association in large scale target that undergo different types of motion in the surveillance area.Focusing on these issues,this paper mainly studies the method of multi-target data association based on measurement sequence classification,attempting to improve the accuracy,continuity and adaptability of target track data association.The specific research contents are as follows:First,the concept of track initiation based on trajectory classification is put forward.The traditional target track initiation problem is regarded as the intelligent classification of the measurement sequences.The data-driven random forests prediction method is employed to cognize target tracking and clutter environment,based on non-parametric decision trees and machine learning theory,improving the processing ability in non-linear and non-Gaussian situation.The training data set is established initially by the system test experiment and simulation and accumulated continuously during radar working.Secondly,two modified methods are proposed to expand the scope of application by analyzing the limitations of the application scope of the cognitive processing method of target track initiation based on track segment classification.There is a delayed initiation or even no initiation for the target with missing alarms.The probability of the detector output is considered by forming the association hypothesis and the missing alarm hypothesis,and the probability of the hypothesis is obtained by the support vector machine based on the posterior probability.Another problem of the algorithm is that it requires a large number of measured data from known sources to construct a classifier model with high accuracy.Semi supervised learning algorithm is introduced to improve the generalization performance of the system by using unlabeled samples.Finally,a radar target track maintenance algorithm based on trajectory classification is proposed.The classification of measurement sequences in the above is achieved through sliding window propulsion,so as to achieve real-time association.By integrating the track initiation and maintenance into the same framework,multiple hypothesis target association system based on trajectory classification can automatically initiate,maintain and delete the track.The experimental results show that the method proposed in this study can effectively solve the target tracking difficulties caused by the lack of prior knowledge,the target motion or the clutter background does not conform to the setting,the data lack,etc.It is of great significance in improving the accuracy and robustness of situation information acquisition and service command and decision-making of radar system.